I spent a couple of weekends running my own photos (and, with permission, a few friends') through online face analysis tools. Not to get a flattering number, but because I wanted to know what these tools actually look at. Here is what I learned, written for anyone who is curious about an attractiveness test but doesn't want to take the score too seriously.
What a face analysis tool really measures
Under the hood, most tools do three things:
- Find landmarks. A detection model places dozens of points on your face: eye corners, brow peaks, the tip and base of the nose, the corners of the mouth, the jawline and chin.
- Turn landmarks into ratios. Distances between points become proportions: face width to face height, eye spacing relative to eye width, the vertical thirds (hairline to brow, brow to nose base, nose base to chin), jaw width compared with cheekbone width.
- Compare the ratios to a reference. That reference is either a classic "ideal proportions" model or a model trained on how people rated a large set of faces.
So when a tool gives you a score, it is really telling you how closely a handful of measurements match its reference. It is not reading your personality, your style, or how you look when you laugh.
Why the same face gets different scores
This surprised me the most. My own score moved by almost a full point between two photos taken a minute apart. The reasons were boring and fixable:
- Lens distortion. A phone held at arm's length (or closer) widens the nose and narrows the ears. Stand back and zoom in slightly, or ask someone else to take the photo.
- Head tilt. Tilting the chin down shortens the lower third and changes jaw width. Keep your head level and look straight at the lens.
- Lighting. Hard side light creates shadows that hide landmarks along the jaw. Soft light facing you works best.
- Expression. A big smile changes mouth width and cheek shape. A relaxed, neutral expression gives the most repeatable result.
- Hair and glasses. Hair over the forehead or a heavy frame can push landmarks into the wrong spot.
If you want a result that means anything, take three photos following those rules and look at the average rather than the best one.
Face shape is the most useful part
Of everything these tools output, face shape turned out to be the most practical. Knowing whether your face reads as oval, round, square, heart, oblong or diamond helps with very ordinary decisions: haircuts, beard shape, glasses frames, even how you part your hair. I used Face Analysis AI for this, and its free face shape analysis gave me a clear answer from a single front-facing photo, along with the measurements it was based on, which made it easy to sanity-check against a mirror and a tape measure.
How to read an attractiveness score without overthinking it
A few rules I settled on:
- Treat it as a measurement of one photo, not of you. Change the photo and the number changes.
- Look at the components, not the headline number. "Lower third slightly long" or "eyes slightly close-set" is more informative than "6.8".
- Focus on what you can change. Grooming, hairstyle, skin care, posture and photo technique all shift results. Bone structure doesn't, and a score isn't a reason to worry about it.
- Remember the reference is narrow. Beauty standards vary by culture and era, and training data has its own biases.
A quick routine if you want to try it
- Face a window in daylight, hair off your forehead, glasses off.
- Have someone take the photo from about 1.5 metres away at eye level.
- Keep a neutral expression and a level chin.
- Run the photo through a face analysis tool and note the face shape and the individual ratios.
- Repeat with two more photos and compare.
If you want more background on photo setup, I found this practical checklist on taking a face analysis photo helpful before I started.
Used this way, an attractiveness test is less a verdict and more a mirror with a ruler attached: handy for haircut and photo decisions, and not worth losing sleep over.
Owen Harte